Approachability, regret and calibration: implications and equivalences
From MaRDI portal
Publication:2438352
DOI10.3934/jdg.2014.1.181zbMath1286.91028arXiv1301.2663OpenAlexW2320006964MaRDI QIDQ2438352
Publication date: 11 March 2014
Published in: Journal of Dynamics and Games (Search for Journal in Brave)
Full work available at URL: https://arxiv.org/abs/1301.2663
2-person games (91A05) Dynamic games (91A25) Rationality and learning in game theory (91A26) Multistage and repeated games (91A20) Online algorithms; streaming algorithms (68W27)
Related Items
Replicator dynamics: old and new ⋮ A general internal regret-free strategy ⋮ Approachability of convex sets in generalized quitting games ⋮ Constrained no-regret learning ⋮ No-regret algorithms in on-line learning, games and convex optimization ⋮ Smale strategies for network prisoner's dilemma games ⋮ Playing against no-regret players ⋮ Approachability in Stackelberg stochastic games with vector costs ⋮ Unnamed Item ⋮ A differential game on Wasserstein space. Application to weak approachability with partial monitoring ⋮ Approachability with constraints
Cites Work
- Unnamed Item
- Unnamed Item
- Unnamed Item
- Unnamed Item
- Unnamed Item
- Unnamed Item
- Unnamed Item
- Unnamed Item
- Unnamed Item
- Unnamed Item
- Unnamed Item
- Unnamed Item
- Unnamed Item
- Unnamed Item
- Approachability of convex sets in games with partial monitoring
- An analog of the minimax theorem for vector payoffs
- Regret minimization in repeated matrix games with variable stage duration
- Exponential weight algorithm in continuous time
- Approachability in repeated games: Computational aspects and a Stackelberg variant
- Non-negative matrices and Markov chains. 2nd ed
- Fiber polytopes
- Subjectivity and correlation in randomized strategies
- Optimal strategies in repeated games with incomplete information
- Learning mixed equilibria
- The weighted majority algorithm
- Calibrated learning and correlated equilibrium
- Potential-based algorithms in on-line prediction and game theory
- Approachability in infinite dimensional spaces
- Regret-based continuous-time dynamics.
- Adaptive and self-confident on-line learning algorithms
- A wide range no-regret theorem
- Regret in the on-line decision problem
- A proof of calibration via Blackwell's approachability theorem.
- Conditional universal consistency.
- An easier way to calibrate.
- No-regret dynamics and fictitious play
- Exponential weight approachability, applications to calibration and regret minimization
- Learning correlated equilibria in games with compact sets of strategies
- Approachability with bounded memory
- Some dimension-free features of vector-valued martingales
- On Martingale Extensions of Vapnik–Chervonenkis Theory with Applications to Online Learning
- A Bound on the Proportion of Pure Strategy Equilibria in Generic Games
- Better-Reply Dynamics with Bounded Recall
- Time Average Replicator and Best-Reply Dynamics
- A Geometric Proof of Calibration
- Repeated Games and Qualitative Differential Games: Approachability and Comparison of Strategies
- Calibration and Internal No-Regret with Random Signals
- Self-Calibrating Priors Do Not Exist
- The Well-Calibrated Bayesian
- Weak Approachability
- Asymptotic calibration
- A Simple Adaptive Procedure Leading to Correlated Equilibrium
- Any Inspection is Manipulable
- Learning Theory
- Consistency of Vanishingly Smooth Fictitious Play
- Excludability and Bounded Computational Capacity
- Stochastic Approximations and Differential Inclusions, Part II: Applications
- Algorithmic Learning Theory
- Internal Regret with Partial Monitoring. Calibration-Based Optimal Algorithms
- Prediction, Learning, and Games
- Learning Theory
- On the Global Convergence of Stochastic Fictitious Play
- A Lipschitzian Characterization of Convex Polyhedra
- A Necessary and Sufficient Condition for Approachability
- Calibration with Many Checking Rules
- Minimax Theorems
- Internal regret in on-line portfolio selection
- A general class of adaptive strategies
- A first course on zero-sum repeated games
This page was built for publication: Approachability, regret and calibration: implications and equivalences
